Vision-Integrated Digital Twin for Industrial Manipulator Unloading

In response to the evolution of industrial manufacturing and the rise of intelligent systems, the demand for real-time simulation and robust performance has become critical. This paper introduces a Vision-Integrated Digital Twin (VIDT) designed for the unloading of industrial manipulators. Initially, the paper elaborates on the VIDT framework, which features a decouplable structure, enhancing its adaptability across various application scenarios. Subsequently, the paper proposes an innovative 3D parcel box pose estimation algorithm. This algorithm, which utilizes an RGB-D camera, advanced deep learning techniques, and primitive geometrics, facilitates real-time vision-guided unloading. The paper concludes with experimental results that validate the efficacy of the proposed pose estimation algorithm and the VIDT.